After the discussion of detection performance and critical constraints of ideal spatial-temporal algorithm for detecting dim point moving targets in background-clutter suppressed IR images under the assumption of independent identical distribution of residual samples containing signal plus noise and nois ealone, a new practically efficient spatial -temporal detection algorithm is introduced in which there is no need knowing priori information about the noise or targets, instead they are estimated from the samples directly. The probabilistic distribution of statistic for binary hypothesis testing is derived and it＇s result is t-distribution function. The detection performance analysis, the problems remained in those two cases and their solutions are also given in this paper.The algorithm is useful for optically detecting meteors, satellites, and other moving targets which move in an arbitrary path across many frames of data.